On 12/13/06, Steve Summit <scs(a)eskimo.com> wrote:
By which I mean, the argument usually boils down to
trying to
decide precisely whether category membership is supposed to
denote an "is-a", "has-a" or "is-related-to" relationship.
But that can't be answered, so the arguments can never really be
resolved, and people have to fall back to using categories not
to implement rigid OO-like inheritable hierarchies, but rather,
looser collections where the only semantic attached to category
membership is "is-kinda-related-to". Some categories and the
Well, yeah, unfortunately this is basically true. We should have a
well-defined category structure, or use really loose and flexible
keywords...we sort of have the worst of both worlds here.
"Category:Arrondissements in Paris" and
"Category:Topics relating
to Paris", rather than just "Category:California" and
"Category:Paris". And we're probably doing a lot of that today.
The downside is it's much easier to guess "Paris" as a category name
than "Topics relating to Paris". What would actually be good would be
stronger, semi-automated systems for funnelling people's guesses into
the right names. There's nothing wrong with someone guessing "Paris"
if a bot can then channel that into the correct name - or at least
highlight it for human recategorisation. Just like we don't "object"
to people adding {{stub}} - we have people that come along and do the
stage two recategorisation.
Wikis work very well when you tolerate this kind of behaviour - don't
ask for perfection up front, but instead appreciate every
micro-improvement.
But there are still (and will always be) lots of
problems when
categories contain other categories, and we'll always be
wondering whether category membership is or isn't or should or
shouldn't be transitive, and it's these larger-scale questions
which we can't (under the current architecture) ever fully
satisfactorily resolve.
Well we've discussed this at length, and I think we can go some way
towards resolving it through either naming conventions or some other
way of formally describing categories. Good rules might be "This
category should not directly contain pages.", "This category is a
geographical hierarchy.", "This category is thematic" etc.
Steve